Method and apparatus for object identification using hierarchical model

1 . An object identification method performed by an object identification system, the method comprising:

detecting an object area from an object image in which an object is located based on a trained object area extraction model and cropping the detected object area by an object identification apparatus;

identifying object type information of an object located in the object area cropped by the object identification apparatus based on an object type inference model by an object management server; and

classifying an object type category of an object located in the cropped object area by the object identification apparatus in a stepwise manner,

wherein the identifying of the object type information identifies object type information using the classified category and the corresponding object type inference model by the object management server.

2 . The method of claim 1 , wherein the cropping of the detected object area detects an object area from the object image based on the trained object area extraction model by using a box surrounding the object or pixels occupied by the object.

3 . The method of claim 1 , wherein the cropping of the detected object area infers an object area by extracting a first feature point through a first image encoder of the trained object area extraction model and extracting a feature point area from the extracted first feature point.

4 . The method of claim 1 , wherein the classifying of the object type category in a stepwise manner classifies the category of an object located in the cropped object area sequentially from a highest major classification to a lowest minor classification by the object identification apparatus.

5 . The method of claim 4 , wherein the identifying of the object type information identifies object type information using the lowest minor category, which is the classified category, and the corresponding object type inference model by the object management server.

6 . The method of claim 4 , wherein the identifying of the object type information determines the object type information as one of a plurality of classes belonging to the lowest minor category by the object management server.

7 . The method of claim 1 , wherein the identifying of the object type information extracts a second feature point through the classified category and a second image encoder of the corresponding object type inference model and infers object type information by entering the extracted second feature point into a feature point decoder by the object management server.

8 . An object identification system using a hierarchical model, the system comprising:

an object identification apparatus detecting an object area from an object image in which an object is located based on a trained object area extraction model and cropping the detected object area; and

an object management server identifying object type information of an object located in the object area cropped by the object identification apparatus based on an object type inference model,

wherein the object identification apparatus classifies an object type category of an object located in the object area cropped by the object identification apparatus in a stepwise manner, and

the object management server identifies object type information using the classified category and the corresponding object type inference model.

9 . The system of claim 8 , wherein the object identification apparatus detects an object area from the object image based on the trained object area extraction model by using a box surrounding the object or pixels occupied by the object.

10 . The system of claim 8 , wherein the object identification apparatus infers an object area by extracting a first feature point through a first image encoder of the trained object area extraction model and extracting a feature point area from the extracted first feature point.

11 . The system of claim 8 , wherein the identification apparatus classifies the category of an object located in the cropped object area sequentially from a highest major classification to a lowest minor classification.

12 . The system of claim 8 , wherein the object management server identifies object type information using the lowest minor category, which is the classified category, and the corresponding object type inference model.

13 . The system of claim 8 , wherein the object management server determines the object type information as one of a plurality of classes belonging to the lowest minor category.